Help a noob design a LED array?

Hey all - this is my first post here. I hope it's not too far off-topic, but in hunting around for info about LEDs, I got quite a few hits here and saw some threads where there were people who clearly knew what they were talking about. So here goes.

I'm working on a bright LED backlight to shine through a 4" LCD panel in use as a homebuild video projector. It's not clear yet if I'll be able to squeeze enough brightness out of this setup to be able to end up with a watchable projection, but it's a fun and inexpensive project, so I'm willing to risk it.

Here is a link to the LEDs I'm using:

http://cgi.ebay.ca/ws/eBayISAPI.dll?ViewItem&item=7578524628

They're white Jeled 5 mm units with a claimed intensity of 40000 mcd.

I'm about 1/2 done mounting and soldering the LEDs, the array will be 18 rows of 14 LEDs, for a total of 252 units. For my testing purposes, I'm using an ATX power supply to drive them, but for the finished version I'd like it if I could switch to a more compact wall-wart type 12V supply.

I'm wiring the rows of 14 LEDs in parallel, and my initial plan was to have 3 "blocks" of 6 rows. Each block would be made up of rows wired together in parallel, and the 3 blocks would be wired in series. From the specs on the page above, these things want about 3.4V and 20 mA each, so this works out to 10.2V and ~1.7A.

My basic plan is to put a variable resistor in series with the array, and tune it until the voltage and current give me the right balance between brightness and longevity.

My questions are:

1) Does this basic approach make sense?

2) With this approach, should I be considering under-driving the LEDS to keep everything safe? Or is there room to over-drive them for better brightness? If so, how much?

3) Are there any benefits to using the ATX power supply in the finished product? It's more expensive and less compact than a regular 12V transformer, but I'd expect the power to be much better regulated.

4) How can I estimate how many lumens I should be getting out of this setup? I'm thinking it should come out close to 1000 based on the reading I've done so far - does that sound about right?

I can post photos later in the week if anyone's interested in this setup.

Thanks a bunch,
Phil.

Hmmm...I don't think this is going to work.

1. A LCD beamer needs a very small light source (point). A LED array is way too large.

2. 1000 lumen is not enough light. I think the transmission ratio of a decent LCD is about 25%, maybe less.

3. I don't think the color spectrum of white LEDs is matched to the filters in the LCD...you'll have to correct this and you will loose light again.

4. Remember to power the array with a constant current source.

5. Watch out for heat..the array will get *very* hot without forced air cooling.

6. Maybe you can try a fancy RGB LED array....this will improve efficiency. However, I still think the light emitting surface is way too large.

7. If you plan to continue the project: good luck! ...and let me know the results!

Hiya, and thanks a lot for the reply.

Wim Hertog said:

Hmmm...I don't think this is going to work.

Click to expand...

Hmmmm... I hope you're wrong, but you may be right.

1. A LCD beamer needs a very small light source (point). A LED array is way too large.

Click to expand...

Well, what it really needs is a collimated light source, which is normally provided by a point source and a collimating lens. My thinking is that the array will provide almost collimated light (due to the relatively narrow beam angle of the LEDs), and it will be the same size as the LCD panel, mounted close behind it.

For projection optics I'm using an HD-6 CRT projection lens which is large enough to accommodate the entirety of the 4" LCD panel.

2. 1000 lumen is not enough light. I think the transmission ratio of a decent LCD is about 25%, maybe less.

Click to expand...

It's quite a bit less than 25%. More like 5-10%. The intended usage for this is strictly darkened room on a medium-sized screen (like 3-4 feet wide), so if I can get around 100 lumens through the LCD panel and optics, I should have something visible, although only time will tell if it's watchable.

This thing is not meant to be a videophile home theatre setup, but basically a toy to experiment with and probably hand off to my little sister as a gift if it works OK.

3. I don't think the color spectrum of white LEDs is matched to the filters in the LCD...you'll have to correct this and you will loose light again.

Click to expand...

That's an interesting issue that I hadn't considered. People using metal halide lamps for these homebuilt PJ's are using all kinds of different lamps with different color temps / spectra / CRI, and I haven't heard reports that the brightness is affected much by these variations. The spectrum of LEDs is a concern in the anti-LED camp, though. I guess we'll see.

4. Remember to power the array with a constant current source.

Click to expand...

This is a big part of what I'm trying to learn more about. A constant current source varies the voltage to provide the same current as the load or resistance changes, correct? Where do I find such a source? Do I look for something adjustable, or try to build a circuit that will supply the 1.7A I need?

Most importantly, will a constant current source be expensive?

And what are the drawbacks of using a simple 12V DC wall converter with a resistor?

5. Watch out for heat..the array will get *very* hot without forced air cooling.

Click to expand...

Will do. So far, I've run as many as 56 LEDs in parallel with the 3.3V output from the ATX power supply for over 30 mins, and they don't get warm to the touch. Does this mean that they're badly under driven in this config, and that I should be able to drive significantly more current through them for more brightness?

6. Maybe you can try a fancy RGB LED array....this will improve efficiency. However, I still think the light emitting surface is way too large.

Click to expand...

For something like this where I basically want a toy, I think that would be too fancy. Worth considering for a future project, but holy crap that would be a lot of soldering.

The 500+ connections I'm soldering with my clumsy fingers are enough of a trial.

7. If you plan to continue the project: good luck! ...and let me know the results!

Click to expand...

I'm definitely continuing, I'll post some pics here later of the LEDs and the projection results if people are interested.

My thinking is that if the PJ is too dim or off-color to be watchable, I'll still have that array for a bright and handy portable worklight powered by something like a 12V lamp battery. Or maybe a quick-connect fitting to hook it up to my motorcycle for camping or roadside adventures.

Thanks again for the reply and advice!

I would suggest a spot MR-16 halogen bulb, something in the 50-75W range and 8 to 12 degrees, you may have to experiment to find the right beam width. Not as durable as the LEDs but deffinately cost effective at $5 a bulb and $20 for a fixture with a 120VAC to 12VDC switch mode power supply.

Hi Syscrush,
I agree with the comments on the size of the array being a problem.....however if you look at something like an overhead projector, where the light source is designed to be a large area and the focussing element projects from the image source....you never know, it might just work, you mentioned your focussing lens is big enough to work the 4inch LCD so...there is goes! Please keep us all informed.

You might want to try a lamina white LED engine, though you'll have heatsinking issues, if the size of the light source becomes a problem.

You didn't mention if the LCD is colour or b/w....the white LEDs would have a problem with colour. Have you considered using the above mentioned RGB array and and b/w LCD..or...splitting your colour image into three quick monochrome images and switching the RGB to do the colour...some of the newer projectors are doing this technique as well as newer LCD TV/monitors are using RGB backlights. Apparently you get very good colour rendition this way. I believe lumileds has patented a scheme for backlighting LCD panels with RGB LEDs. Also the projectors are using a mirror chip rather than an LCD panel to improve transmission (or reflection I suppose) efficiency.

In terms of heat, many LED luminare designers are looking at using 5mm LEDs for this reason. It's still an issue, but less of an issue than high power LEDs.....have a look at the december issue of Ledsmagazine for an article on this sort of thing.

www.instructibles.com had an article on a guy designing a white LED bike light array.....unfortunately at the time I'm writing this the site was down. When it comes up again I'll try to find the link. This would be a good site to post your project when it's finished.

Funnilly enough I'm doing a little 24x100mA 10mm LED RGB array at the moment myself, for a wide angle wash light....will post when it gets first light. From previous work, I've found it's important to keep an alternating RGB pattern around the edges of the array or else you end you with colour fringing.

Phil

Syscrush said:

This is a big part of what I'm trying to learn more about. A constant current source varies the voltage to provide the same current as the load or resistance changes, correct? Where do I find such a source? Do I look for something adjustable, or try to build a circuit that will supply the 1.7A I need?

Most importantly, will a constant current source be expensive?

And what are the drawbacks of using a simple 12V DC wall converter with a resistor?

Click to expand...

For LEDs we need current regulators? You have the choice between a simple and cheap linear regulator or a more expensive switching regulator.

Linear regulators (the LM317 IC is very famous) just "burn off" excess voltage...in other words, they get hot and are inefficient.

Switchers are switching the current with transistors through an inductor to regulate. They can be very efficient (>90%) but are more complex and expensive.

If I were you, I would use the LM317 IC. Do a google search for "LM317 current regulator".

Syscrush said:

Will do. So far, I've run as many as 56 LEDs in parallel with the 3.3V output from the ATX power supply for over 30 mins, and they don't get warm to the touch. Does this mean that they're badly under driven in this config, and that I should be able to drive significantly more current through them for more brightness?

Click to expand...

My DIY bikelight runs 48 very efficient white LEDs (the famous 26Kmcd LEDs) at a max of 30 mA/piece. I can guarantee you that the PCB gets too hot on full power when I'm not moving. Also, never use just a voltage regulated power supply to drive a large LED array.

Syscrush said:

I'm definitely continuing, I'll post some pics here later of the LEDs and the projection results if people are interested.

My thinking is that if the PJ is too dim or off-color to be watchable, I'll still have that array for a bright and handy portable worklight powered by something like a 12V lamp battery. Or maybe a quick-connect fitting to hook it up to my motorcycle for camping or roadside adventures.

Thanks again for the reply and advice!

Click to expand...

We're always interested in pictures!

rgbphil said:

Hi Syscrush,
Funnilly enough I'm doing a little 24x100mA 10mm LED RGB array at the moment myself, for a wide angle wash light....will post when it gets first light. From previous work, I've found it's important to keep an alternating RGB pattern around the edges of the array or else you end you with colour fringing.

Phil

Click to expand...

Hehe, me too I just ordered my red, green, blue and white superflux LEDs....50 pieces/color. The plan is to transform a halogen floodlight into a RGBW color washer. A PIC µC will take care of the color mixing.

I'm also going to study the color rendering of RGB white. We had an interesting disussion a while ago about the fact that colors stand out much more in RGB white than in high CRI white from fluorescent lamps (despite the fact that RGB white from LEDs has a very low CRI).

I'll post some pics when everything's finished (will take some time).

Hi Wim,
Here's the bits for what I'm intending to do....a fLEDlight:



Were you thinking of something similiar.....this one is actually being designed around the PSU, I got a bunch of these little Kodak camera chargers for about $AU2.50 from Jaycar in their odds and ends bin, 7V at 2.1A...perfect. The green/blue LEDs will be stringed in twos, and the reds in threes (except for the odd one out).

You've probably seen this one before in my previous shameless promotional posts....but here is a another wash light, an eLEDtric radiator.

(five channel RGB pattern/sequencing, IR remote controlled jobby).

Incidentally, you mentioned never using a constant voltage regulator for arrays.....could you list the reasons please.

I was intending to do a very simple cct using ULN2803s to control the strings and dropping resistors, not wanting to spend too much time on the project. A simple PIC to do the colour control. As in anything there are compromises, so far I can see:
- variations in light output due to variations in the LED forward voltage. I can handle that with this application, there is no need to evenly light say a backlight panel. Perhaps in Syscrushs application it WILL be important for even illumination across his panel to get a good picture.
- ineffiency+heat due to wastage in the dropping resistors. Well a constant voltage+resistor is effectively the same as a voltage regulator like the LM317. By keeping the voltage close to the forward drop of the LED string+switching transistor we can reduce the voltage drop across the resistor, hence efficiency isn't too bad. With the 7V supply it's all very close.
- variations in the PSU possibly overdriving the LEDs on occasion, particularly during colour switching...I compromise here by underdriving the LEDs by about 10-20% and making up less light output with more LEDs....I'm never happy about using components at full ratings anyway.

Obviously a constant current cct is best all around, but I didn't see the need for that in this quicky cct. I'd still like to hear your thoughts on your statement about 'never' using a constant voltage source.

Phil

Incidentally, here is the instructables link
http://www.instructables.com/ex/i/A9D07B9C83FD1028A786001143E7E506/
cool site!

OK - I'm definitely in the right place.

The spec in the page I ordered these LEDs from show the typical values at 20 mA, and the spec for the max continuous current is 20-25 mA. So I want to source or design a power supply that will provide that for each LED, however I end up wiring this all together.

But let's assume that I ignored that advice (being the ******* that I am) and stuck with the ATX power supply driving all 252 LEDs in parallel off the 3.3V output. What's the risk? Just reduced efficiency, or is there danger of burning out the LEDs or the power supply?

I'm also stoked to see mention of these RGB washlight arrays, as a big part of my motivation in doing this project is to learn about LED stuff in a simple context before I make some multi-channel multi-color lighting pieces for my home.

Thanks again for all the insight and advice,
Phil.

Hi Syscrush,

I don't see any problem running 20mA LEDs off your supply, except for all the limiter resistors! Assuming forward voltage of 3.2V or so at 20mA thats only 64mW per LED or 252*64 = 16.1W total. Check the supply can provide at least 20% more....I'd imagine it would.

To the heat issue:
If each LED occupies say a 5mm radius (including spacing) that'd be 252 * pi * 0.005*0.005 = approx 0.2m^2 (actually a bit more because of the I presume hexagonal packing). So you'll have 16.1/0.2=80W/m^2 to dissapate..easy.....you should be able to do this by designing the PCB with big copper areas for the cathodes top and bottom and possibly mounting these copper areas to the framework as a further measure. Fiddle with the numbers depending on what your actual packing density is.....and work out the hex packing area properly.

Think lots of copper areas, rather than thin tracks, where the copper goes the heat goes. If you can move it to the frame and away, even better. Make sure the PCB can accomodate the addition of some heatsinks (even bent bits of sheet copper will do, or angle bracket aluminium), if you have problems with heat, just put heatsinks on the back. It's important to use big copper areas for the cathodes, because this is where most heat comes out, the LED die is usually mounted with the cathode on the die cup and a thin wire to the anode. However do check this, I've seen LEDs with anodes mounted to the die cup.

I do have a suggestion however, and that is to use the 12V supply and string together 3 or 4 LEDs in series instead. That'd save lots of limiting resistors.

OK, here are some pics of the work in progress:

First, the array from the top. More about the odd profile of these LEDs below:


Here you can see how each row of 14 LEDs is wired in parallel:


Close-up of my weak soldering skills:


Now, to get the LEDs packed as tightly together as possible, I ground off the little lip that goes around the base. It's not a ton of work (especially compared to the mounting and soldering!), but it makes a huge difference in how many LEDs I can fit in the 8 square inches behind my LCD panel. So here's a before and after shot:


And how tightly can I pack the LEDs after making this little mod? VERY tightly:


The total area will not be anywhere near 0.2 square meters. It'll be more like 0.0063 square meters, or about 8 square inches.

When I'm done mounting and soldering, I'll have 18 rows of 14 LEDs each wired in parallel. My plan is to split those 18 rows into 3 blocks of 6 rows each (for a total of 84 LEDs per block). All 84 LEDs in each block would be wired in parallel (for a 3.4V 1.7A array), and then those 3 blocks would be wired in series (for a 10.2V 1.7A array). Then I can use the 12V output of my ATX power supply or possibly a "wall wart" type 12V transformer and a variable resistor to drive it.

If heat's an issue, I thought maybe I could cover the back with heat sink grease and just put a big aluminum CPU cooling fin block on top to dissipate the heat.

Does this have any chance of success? I am under the impression that strictly speaking I should have a resistor in series with each individual LED, but clearly that's not happening here. One per block or even one per row is doable, but 252 is out of the question.

Note that I have done 1 test where I ran 15 LEDs all wired together in parallel hooked up straight to the 3.3V output on the ATX PSU and let it run for over 30 mins. No indication that the LEDs were doing anything funky, and they never got warm to the touch.

Thanks again for all the insight and pointers so far, and thanks in advance for all of the "OMG don't do that, you're gonna blow something up!" comments that are surely on their way.

Take it easy,
Phil.

Crikey.....that's a lot of LEDs! You don't need a resistor in series with each LED at all....I'm a bit confused by your block/parrallel/series wiring description.....each little string will need to be electrically seperate...is that what you mean. Either way at 12V supply you'd have 4 max leds in a string (under driving them a bit)....
Anyway, looks pretty good. I doubt you'll have heat issues, but with the wiring the way it is, if you do there would be no way to add a heat sink. Peeerrrrhapppssss you could try a silicone heatsink sheet pressed up against the LED wiring to conduct heat to a plate behind without shorting anything....a bit iffy though. Still if you're not having problems, no need to complicate things.
The problem with parrallel connections is that one LED may 'hijack' the forward voltage of the rest of them and suck most of the current....leading to one getting very bright and the others dimmer....eventually that LED with the lowest forward voltage may burn out if the voltage isn't carefully controlled....I'd be parralleling strings+limit resistors, not individual LEDs to avoid this.
Phil

rgbphil said:

Crikey.....that's a lot of LEDs!

Click to expand...

Yeah, and it's just over 1/2 done. 140 of 252 LEDs ground, mounted, and soldered in place so far.

You don't need a resistor in series with each LED at all....I'm a bit confused by your block/parrallel/series wiring description.....each little string will need to be electrically seperate...is that what you mean.

Click to expand...

OK, each block is 84 LEDS all wired together in parallel. From the pics you can see that each row is composed of 14 LEDs wired together in parallel, and I'm planning to tie all the +ve terminals for 6 rows together, and all the -ve for six rows to give me my 84 LED block.

Each block of 84 LEDs will want 3.4V driving a total of about 1.7A (20 mA x 84 LEDs). I have 3 blocks of 84 LEDs each, and these three blocks are wired together in series, so it's like 3 mungo-big 3.4V 1.7A LEDs wired together in series for 10.2V 1.7A total.

It's hard to explain this in words and too late to draw a diagram. If it's still not clear, then I can draw something up tomorrow.

Thanks very much for the explanation of how a single LED with a lower forward voltage can ruin the party. That makes a lot of sense. With so many LEDs wired together in parallel, I guess this design is particularly susceptible to that problem. Hmmmm...

Thanks again,
Phil.

Wim Hertog said:

Hmmm...I don't think this is going to work.

1. A LCD beamer needs a very small light source (point). A LED array is way too large.

Click to expand...

It can be done if you focus on the the LCD, been there, done that.

Wim Hertog said:

2. 1000 lumen is not enough light. I think the transmission ratio of a decent LCD is about 25%, maybe less.

Click to expand...

Try 4%, and up to 7.5% (nearly double the light) if you use 3M DBEF or a reflective polarizer immediately behind the LCD.
http://products3.3m.com/catalog/us/...ikuiti_3_0/command_AbcPageHandler/output_html

Also consider using BEF films, which will "steer" the light into a cone towards your lens, instead of lighting up areas where your lens is not. A crossed set of these films will give you another 111% increase in brightness:
http://products3.3m.com/catalog/us/...ikuiti_3_0/command_AbcPageHandler/output_html

Wim Hertog said:

3. I don't think the color spectrum of white LEDs is matched to the filters in the LCD...you'll have to correct this and you will loose light again.

Click to expand...

When using white LEDs, the reds will be severely lacking, and have a pinkish white color, the greens will look a bit washed out also. The white backlight will turn a greenish hue, that can be offset a bit with the right binning choice of white.

Wim Hertog said:

5. Watch out for heat..the array will get *very* hot without forced air cooling.

Click to expand...

Removing the heat will be important for maximum light output as well as LED life.

DCFluX said:

I would suggest a spot MR-16 halogen bulb, something in the 50-75W range and 8 to 12 degrees, you may have to experiment to find the right beam width. Not as durable as the LEDs but deffinately cost effective at $5 a bulb and $20 for a fixture with a 120VAC to 12VDC switch mode power supply.

Click to expand...

If you shine 50-75W at the LCD, you will probably clear it due to the temperatures. An alternative is to use a dichoric heat reflector, or heat absorbing glass and a fan. Typically when using this approach, folks will use a fresnel lens, in an attempt to fix the issue with a point light source.

rgbphil said:

Anyway, looks pretty good. I doubt you'll have heat issues, but with the wiring the way it is, if you do there would be no way to add a heat sink. Peeerrrrhapppssss you could try a silicone heatsink sheet pressed up against the LED wiring to conduct heat to a plate behind without shorting anything....a bit iffy though.

Click to expand...

I was thinking I could make kind of like a little "sandbox" around the wiring on the back and fill it with thermal grease to conduct the heat to the heat sink. I'd have a layer like 2-3mm thick to clear the wiring. If the thermal grease is electrically conductive, then I could varnish the wiring first before loading it up with the grease and attaching the cooling fin block.

What's the limit on how thick you can trowel on the thermal grease and still have it be effective? God, it occurs to me now that I'd have to do something too to keep the grease from oozing through the holes in the perfboard. Ugh.

This is only if temps are a problem, of course. I haven't put an ammeter on this setup yet when running large #'s of LEDs, so my quickie tests may have been under-driving by a lot.

Syscrush said:

I was thinking I could make kind of like a little "sandbox" around the wiring on the back and fill it with thermal grease to conduct the heat to the heat sink. I'd have a layer like 2-3mm thick to clear the wiring. If the thermal grease is electrically conductive, then I could varnish the wiring first before loading it up with the grease and attaching the cooling fin block.

What's the limit on how thick you can trowel on the thermal grease and still have it be effective? God, it occurs to me now that I'd have to do something too to keep the grease from oozing through the holes in the perfboard. Ugh.

This is only if temps are a problem, of course. I haven't put an ammeter on this setup yet when running large #'s of LEDs, so my quickie tests may have been under-driving by a lot.

Click to expand...

18x14 LEDs = 252 LEDs
3.3Vf*20mA= 0.066 Watts each
252 LEDs * 0.066 Watts each = 16.632 Watts total.

IMHO, you do need heatsinking in this small of a space with that many LEDs and that amount of Watts.

Watch out for things heating up and the LED Vf's dropping, and pulling even more current, and making yet more heat. Thats part of what a constant current source will help you with.

You might consider a thermal gap pad from Berquist, Dow Corning, and many others:
http://www.dowcorning.com/content/etronics/etronicspadsfilm/default.asp
http://www.dowcorning.com/content/etronics/etronicspadsfilm/etronics_tcpad_gapov.asp

There are also a variety of thermally conductive epoxies and silicone "glues" on the market, but reworking would be a pain.

In case you missed it, see my post earlier.

NewBie said:

It can be done if you focus on the the LCD, been there, done that.

Click to expand...

Cool. With an LED light engine?

Also consider using BEF films, which will "steer" the light into a cone towards your lens, instead of lighting up areas where your lens is not.

Click to expand...

I'm using a CRT projection lens where the diameter is slightly larger than the LCD panel, and mounted about 1" away from it. Any light that makes it through the panel *should* hit the projection lens - there should be no requirement to steer it into a cone.

When using white LEDs, the reds will be severely lacking, and have a pinkish white color, the greens will look a bit washed out also. The white backlight will turn a greenish hue, that can be offset a bit with the right binning choice of white.

Click to expand...

Thanks for helping me set my expectations.

If you shine 50-75W at the LCD, you will probably clear it due to the temperatures. An alternative is to use a dichoric heat reflector, or heat absorbing glass and a fan.

Click to expand...

I have some IR glass in case I need it.

Typically when using this approach, folks will use a fresnel lens, in an attempt to fix the issue with a point light source.

Click to expand...

This design shouldn't need any fresnels - in principle, I've got a collimated light source behind the LCD (so no need for a condenser lens), and the size of the projection lens is bigger than the LCD and too close to it for the light to diverge out of the lens' field of view (so no need for a field lens).

Thanks!

NewBie said:

IMHO, you do need heatsinking in this small of a space with that many LEDs and that amount of Watts.

Click to expand...

Noted. Thanks.

Watch out for things heating up and the LED Vf's dropping, and pulling even more current, and making yet more heat. Thats part of what a constant current source will help you with.

Click to expand...

I had a feeling that this scenario was the issue. Thanks.

You might consider a thermal gap pad from Berquist, Dow Corning, and many others:

Click to expand...

Cool! Thanks for the pointers!

There are also a variety of thermally conductive epoxies and silicone "glues" on the market, but reworking would be a pain.

Click to expand...

Yeah, but reworking is the LAST thing I'm gonna want to do. I was thinking of using potting compound (or similar) to encase/protect the wiring in the back anyhow. Maybe with the right choice of epoxy I can kill 2 birds with one stone? Of course, the pad mentioned above would leave room for repairs if needed.

Thanks again!
Phil.